Quantile Regression for Longitudinal Data with Left-Censoring and Informative Dropouts

• Activity Number: 139
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Minjae Lee*+ and Lan Kong
• Companies: University of Pittsburgh and University of Pittsburgh
• Keywords: Biomarker ; Detection limits ; Informative dropout ; Left-censored data ; Longitudinal data ; Quantile regression

Abstract

Biomarkers are often measured repeatedly in biomedical studies to help understand the development of diseases. In the Genetic and Inflammatory Markers of Sepsis (GenIMS) study, the biomarker data were collected during the course of hospitalization for patients with community acquired pneumonia. However, the longitudinal analysis of biomarkers is complicated by informative drop-outs due to death or discharge early and left censoring due to detection limits of assays used. To account for these two issues, we consider a weighting technique for quantile regression models which impose minimal assumptions on the distribution of the data. In particular, we weight the estimating equation for censored quantile regression by the inverse probability of drop-out. We evaluate our method through simulation studies and use GenIMS data set for demonstrations.